Two groups have recently discovered a statistically significant deviation in the fluxes of high-redshift type Ia supernovae from the predictions of a Friedmann model with zero cosmological constant. In this Letter, I argue that bright, dusty, starburst galaxies would preferentially eject a dust component with a shallower opacity curve (hence less reddening) and a higher opacity/mass than the observed galactic dust which is left behind. Such dust could cause the falloff in flux at high-z without violating constraints on reddening or metallicity. The specific model presented is of needle-like dust, which is expected from the theory of crystal growth and has been detected in samples of interstellar dust. Carbon needles with conservative properties can supply the necessary opacity, and would very likely be ejected from galaxies as required. The model is not subject to the arguments given in the literature against grey dust, but may be constrained by future data from supernova searches done at higher redshift, in clusters, or over a larger frequency range.